High-Throughput Flow Field & Crowd Pathfinding Engine

Pure Python flow field pathfinding with 127K+ agent evals/sec for 25,000 concurrent agents, spatial-hash avoidance, binary .flow serialization, and 312 tests.

Production-grade, headless Flow Field (Vector Field) pathfinding and crowd simulation engine implemented entirely in pure Python standard library. Features prioritized multi-goal Dijkstra integration wavefronts (613K cells/sec), gradient vector derivation (396K cells/sec), bilinear continuous velocity sampling, spatial-hash soft avoidance forces and hard push-out, 25,000+ concurrent agents with flocking at 127K evals/sec, high-density 16-byte binary .flow format, full JSON export, and 312 unit tests passing in 1.45 seconds.

• 127,933+ Agent Evaluations/Sec with 25,000 Concurrent Agents
• 613,000+ Cells/Sec Dijkstra Wavefront Integration Field
• 396,000+ Cells/Sec Normalized Vector Field with Bilinear Sampling
• Spatial-Hash Collision Avoidance & Push-Out (0.7 Radius Tuned)
• 312 / 312 Automated Pytest Tests Passing (100% Deterministic)
• High-Density 16-Byte Packed Binary .flow Format + JSON Exporter
• Zero External Dependencies (Pure Python Standard Library)
• 100% Royalty-Free Perpetual Commercial Rights